The argument against this approach is that it's not always humans accessing web pages. And even if it is humans, sometimes they need computer assistance to interpret the content.
http://en.wikipedia.org/wiki/Semantic_HTML
As the article says, this is equivalently achieved by naming div classes and IDs (or some other attribute like itemprop a la schema.org). Also I'd say a completely style-free HTML scheme with open-ended semantic tag names (XML) styled by CSS is equivalent. The point is HTML doesn't serve a useful styling purpose.
I don't have a strong opinion on whether browsers should provide default styling, as it's trivial to override (although many web pages do rely on it). I agree that HTML tag names don't serve a very useful styling purpose, but they provide a useful structural purpose that cannot be easily replaced by using open-ended tag names or classnames. Custom Elements (http://w3c.github.io/webcomponents/spec/custom/) are in the process of being implemented, but I don't think that's an argument for throwing out the standard, structural elements that provide a common API for interpreting document structure.
